This feasibility study explored the automated data collection, scoring, and reporting of children's complex problem-solving processes and performance in Web-based information-rich environments. Problem solving was studied using realistic problems in realistic contexts demanding multiple cognitive processes in the domain of environmental science. Sixty-nine middle school and high school students completed pretest and posttest concept maps, a relevant bookmarking measure, and a metacognitive survey. Process data was collected using computer trace data. During posttest map construction students had access to a simulated World Wide Web environment. The relevant bookmarks students used were web documents, specific universal resource locators, that helped students make at least one meaningful link between two or more concepts. Students were trained in the computer-based concept mapping software and then were given 20 minutes to construct a closed concept map about environmental science. After a tutorial in the bookmark application and the simulated Web environment, students were instructed to search the simulated Web for 40 minutes to help them improve their maps. Data are not complete for all participants due to computer crashes and data log management problems. Student performance from pretest to posttest for concept mapping scores did improve, but more exploratory information seeking behavior did not predict higher scores. Extracting relevant information from the Web resulted in higher final concept mapping scores, but high scores on metacognition did not predict high scores on the relevant bookmarks and final concept maps. There were no significant results for the use of feedback contributing to higher final concept mapping performance or a greater number of relevant bookmarks. (Contains 43 references.) (SLD)